Trained on Anything v3 with its respective vae

Images were generated using a blend of Anything v3 and AOM2-Hard with Anything v3 vae

Highly recommend using a good style LoRA in combination with this. Otherwise, in my experience, it could generate bland and flat images.

I recommend 0.7 - 1.0 but it could go lower.

For her uniform, use the tag IchikaUniform to activate it.

I might end up using a larger dataset to train. I only used around 25 images to train instead of 40+ images like usual.

Check out my other Ano Natsu de Matteru characters!

Kanna Tanigawa | 谷川 柑菜

Styles used:

Dohna Dohna

Afrobull

I hope you enjoy!